Great to see other videos focused on what an ML engineer’s day to day is like, technically. I work in the industry as an MLE myself, and it’s always hard explaining to people what that entails. For me it’s a combination of data science and software engineering, where you need expertise of machine learning concepts and how to apply them to the business objective, but you also need to understand how to build software and infrastructure to serve the models. It’s a complex, interdisciplinary task, and part of why it’s so difficult to get into. Knowing everything from ETL and data pipelines to statistics and probabilistic insight, to backend software development is done through experience in trial by fire, and not learnable in a course. Great video!
Great content! Having recently settled into a workflow+stack myself, I really appreciate this kind of tutorial, one that is holistic in its approach, presenting a full professional toolchain rather than a single technology isolated. Keep it up and good luck!🎉
I'm really liking these kind of long videos, where you combine all ML engineering pieces together! I got a question: in the training unit tests you could use something like the Hypothesis Python Package to create a mock dataset instead of preparing one for that, right? Again, great work!!!🤗
This series was super fun, glad you liked it. I took a peek at Hypothesis and you can use it to feed in test data to your unit tests. So rather than - say - manually setting a number or a string to test - you would just say "run this test on various strings" using the text Hypothesis strategy. Pretty neat. Consider that in the ML context we are concerned with testing only data that might appear in the training/test data
Great to see other videos focused on what an ML engineer’s day to day is like, technically.
I work in the industry as an MLE myself, and it’s always hard explaining to people what that entails. For me it’s a combination of data science and software engineering, where you need expertise of machine learning concepts and how to apply them to the business objective, but you also need to understand how to build software and infrastructure to serve the models. It’s a complex, interdisciplinary task, and part of why it’s so difficult to get into. Knowing everything from ETL and data pipelines to statistics and probabilistic insight, to backend software development is done through experience in trial by fire, and not learnable in a course. Great video!
Great content! Having recently settled into a workflow+stack myself, I really appreciate this kind of tutorial, one that is holistic in its approach, presenting a full professional toolchain rather than a single technology isolated. Keep it up and good luck!🎉
I'm really liking these kind of long videos, where you combine all ML engineering pieces together!
I got a question: in the training unit tests you could use something like the Hypothesis Python Package to create a mock dataset instead of preparing one for that, right?
Again, great work!!!🤗
This series was super fun, glad you liked it. I took a peek at Hypothesis and you can use it to feed in test data to your unit tests. So rather than - say - manually setting a number or a string to test - you would just say "run this test on various strings" using the text Hypothesis strategy. Pretty neat. Consider that in the ML context we are concerned with testing only data that might appear in the training/test data